|
 |
Bimonthly Since 1986 |
ISSN 1004-9037
|
|
 |
|
|
Publication Details |
Edited by: Editorial Board of Journal of Data Acquisition and Processing
P.O. Box 2704, Beijing 100190, P.R. China
Sponsored by: Institute of Computing Technology, CAS & China Computer Federation
Undertaken by: Institute of Computing Technology, CAS
Published by: SCIENCE PRESS, BEIJING, CHINA
Distributed by:
China: All Local Post Offices
|
|
|
|
|
|
|
|
|
|
|
09 May 2023, Volume 38 Issue 3
|
|
|
Abstract
The Multiply-Accumulate (MAC) processing can be implemented using an approximative computing technique. Unlike earlier research, where the approximate range was constrained by an error accumulation issue. In contrast, a number of approximation multipliers are interleaved in this study to account for mistakes that occur when accumulate operations are performed in the opposite direction. Create the roughly 4-2 compressors that produce mistakes in the opposite direction while first reducing the computational expenses for the balanced error accumulation. Then, to provide a comparable error distance, positive and negative multipliers are painstakingly developed based on the probabilistic analysis. According to simulation results on various real-world applications, the suggested MAC processing offers the energy-efficient computing scenario by widening the range of approximatively portions. The core-level energy is decreased by the proposed interleaving strategy.
Keyword
Convolutional Neural Networks, Multiplier, Image Processing, Approximate Computing,Multiplier.
PDF Download (click here)
|
|
|
|
|